THE BEST METHOD FOR PRESENTATION

OF RESEARCH RESULTS
Veljko Milutinovic
emilutiv@etf.bg.ac.yu
(381-11) 762-214

Department of Computer Engineering

School of Electrical Engineering
University of Belgrade
POB 816
11000 Belgrade
Yugoslavia

Abstract (d) Syntax of references;

The major goal of this paper is to serve as a guideline
for organization of research presentations in oral or
written form. Another important goal of this paper is to
convince the researchers to use the author's semantics-
based layout strategy for transparencies. The major
purpose of the entire effort is to make the research
presentations as easy to comprehend as absolutely
possible. Proper usage of the guidelines and strategies
defined in this paper is a conditio sine qua non for those
graduate students who have chosen that the author of
this paper be their major professor. The same structure is
being used for thesis work, as well as for conference and
journal publications, or technical reports to research
sponsors, both by graduate students and professional
engineers.
1. Introduction
This paper focuses on a method for presentation of
research results (in written and/or oral form) and focuses
on the following issues:
(a) Selection of the title;
(b) Structure of the abstract;
(c) Structure of the figures and/or tables
and their captions;
(e) Structure of the written paper
and the corresponding oral presentation
using transparencies;
(f) Semantics-based layout of transparencies
for an oral presentation.
Intentionally, the entire text to follow has been made
relatively short, so more people decide to read it. This
paper represents the decades-long research experience of
the author, and summarizes the mandatory requirements
that he places before his graduate students.
The motivation to publish this paper (which is in use
at the University of Belgrade for about half decade now)
came after the repeated pattern at international
conferences where lots of good research was presented in
such a way that research results are obscured by poor
presentation. It was not possible to understand quickly,
either the essence of the contribution, or the most
important research details.
At a recent major set of computer science/engineering
conferences, no single presentation was following either
the guidelines presented below, or the semantics-based
layout of transparencies to be defined below.
2. Selection of the Title
The selection of title should be both didactic and
lapidaric.

IEEE TCCA NEWSLETTER, September 1996. 1

 In this context, didactic means creating a title which
enables an expert to figure out the essence of the basic
idea and the main contribution, even without reading the
paper; lapidaric means creating a title which induces the
reader to think deeply over the "philosophy" of the
contribution described in the paper.
A relatively good example of a didactic and lapidaric
title is:
APPLYING ENTRY AND LAZY RELEASE
SELECTIVELY:
TEMPORAL VERSUS SPATIAL DATA
This title is didactic since it is immediately obvious
that the main idea is to apply the entry consistency
model to temporal data and the lazy release consistency
model to spatial data, for the performance which is better
than applying only one of the two models to all data.
This title is also lapidaric, since one immediately
starts thinking about how the selective application of two
different consistency models was really implemented.
An alternative (bad) title would be:
SOME ISSUES
IN MEMORY CONSISTENCY MODELING
People would tend to stay away from a paper with
such a title, since that kind of title might be viewed as an
introduction into a contents-free paper, unless it comes
form a well known expert who has a reputation of
knowing what he/she is doing. Consequently, a good
idea may not be noticed by the research community, and
those who reinvent it at a later time will get the credit
instead of the initial inventor.
 6WUXFWXUHRIWKH$EVWUDFW
Wherever possible, the abstract of a research paper
should include the following five elements:
(a) Problem statement of the research under
consideration;
(b) A short list of existing solutions and what is
their drawback, from the point of view of the
above defined problem statement;
(c) Essence of the proposed solution, and why it is
expected to be better under the same conditions;
(d) What type of analysis was done to show that the
proposed solution is really better than any of the
existing ones, from both the performance and
IEEE TCCA NEWSLETTER, September 1996.
the complexity points of view (if one is an
engineer, then both performance and complexity
are equally important);
(e) What are the major numerical highlights of the
analysis (if one is an engineer, numbers are the
"name of the game").
If a 50-word abstract is required, then each part above
should be about one sentence long; if a 500-word
abstract is required, then each part above should be about
10 sentences long, etc. Of course, the language should be
simple and concise, with declarative sentence structure,
written primarily in the present tense.
3. Structure of the Figures and/or Tables
and the Related Captions
Figures and tables should include only language-
independent mnemonics (derived from English
language), which is especially important for non-
English-speaking researchers, and for those writing for
many languages, so it is easier to switch back and forth
between languages.
All details must be clearly visible, even after the same
figure is ported to a transparency for an oral
presentation.
Captions deserve a special attention, which is
neglected in a typical written presentation. The main
issue is that reading only the figure captions of the paper
can substitute the first rough reading of the entire paper.
This goal is achieved more successfully if the caption
includes the following five elements:
(a) Title with the main highlight, i.e. the main
issue to be demonstrated by the corresponding
figure/table;
(b) Legend, to explain all language-independent
mnemonics inside the figure/table;
(c) Description, of one or more phenomena which
deserve attention (e.g., curves A and B cross
each other at X=16);
(d) Explanation, of the essential reason for such a
behavior (e.g., the curves cross each other
because for higher values of X, the following
happens ...);
(e) Implication, or what is to be kept in mind when
designing/engineering a system to exploit the
above noticed phenomenon (e.g., increasing the
size of register file helps until the number of
registers reaches a critical value; after that ...).
2
 A book which insists on this type of reasoning is
[Flynn95]; however, the approach has not been
formalized, and this type of reasoning can not be found
in figure/table captions. Writing a good caption of this
type is extremely difficult for the one who writes the
paper (and graduate students often show resistance to
such an approach), but extremely useful for the one who
reads the paper (and readers/reviewers often show
appreciation for such an approach).
Also, this type of caption may become relatively long,
and one might think that the limited paper space is not
used rationally; however, the captions should include
only the facts which are "local" to the figure/table, and
these facts should never be repeated again in the main
body of the paper. The main body of the paper should
include only the "global" facts (e.g., comparing the
findings from different figures, and similar).
A similar approach can be found in the famous books
of Hennessy and Patterson (alphabetical order), except
that their captions do not always have all five elements,
and if they do include all five elements, these elements
are not formally separated, which is a requirement of the
methodology presented here.
All figure and figure captions should be completed
before the actual writing of the paper starts.
4. Syntax of References
This is another item to be completed before the
writing of the paper starts. As far as the syntax of
references, it is most natural that one follows the syntax
used by the most prestigious scientific journal in the field
(e.g., IEEE Transactions on ...).
If an alternative approach seems to be better, this
methodology suggests that one waits until the major
journal accepts it.
As far as the method of pointing to a reference, the
mnemonical approach with the entire name of the first
author and the year is preferred (so the reader knows
immediately what research group the paper comes from).
Often, the name of the last author conveys that
information more clearly, but it is not practical to use it,
when pointing to a reference. Of course, if so required,
the above method can be easily converted into the
numeric form, mandatory in some journals.
An important reason for doing references before the
actual writing starts is that one makes sure that no
important reference is omitted; a task more difficult to do
after the entire paper is completed.
IEEE TCCA NEWSLETTER, September 1996.
5. Structure of the Written Paper and
the Corresponding Oral Presentation
In the case of a research paper, whenever possible,
one should first develop the skeleton of the
paper/presentation, to include the following first level
titles:
(a) Introduction, to include the basic facts needed
to tune the reader to the paper and/or
presentation;
(b) Problem statement, to define precisely the
problem being attacked by the research under
consideration, and why is that problem
important;
(c) Existing solutions and their criticism, to
survey briefly the major existing solutions form
the open literature and to underline their
deficiencies from the point of view of interest
for this research, which is defined in the above
mentioned problem statement section;
(d) Proposed solution and why it is expected to be
better, to give the essence of the proposed
solution (i.e., the essence of the idea which is to
be introduced), followed by a logical and/or
philosophical discussion about the expected
benefits stemming from the idea;
(e) Conditions and assumptions of the research
to follow, to summarize the environment of
interest. The term conditions refers to the
specifiers of the real environment, and the term
assumptions refers to the simplifications which
simplify the analysis without any negative
impacts on the validity and representativeness
of the final results. It is useful for the reader if
conditions and assumptions are itemized (e.g.,
application-, system-software,- architecture-,
organization-, design-, and technology-related);
(f) Analytical analysis, to show one or more of the
following:
(f1) proof of validity of the major idea of the
paper/presentation;
(f2) calculation of initial values for simulation
analysis to follow;
(f3) rough estimation of the performance;
(f4) rough estimation of the complexity;
(f5) something else which is relevant;
Analytical analysis will not give the final
answers; however, it will help understanding
3
 the concept (it will be helpful both to the
researcher and the reader);
(g) Simulational analysis, to show performance
(this should be the major and the longest part of
the paper);
(h) Implementational analysis, to show complexity
(for some types of research, this one could be
the major and the longest part of the paper);
(i) Conclusion, with the following three major
elements:
(i1) revisiting the major contribution
from the performance/complexity point of view;
(i2) stating who will benefit
from the presented results;
(i3) what are the newly open problems
and research avenues.
One should keep in mind
that some people read only the abstract
and the conclusion;
(j) References, as described above.
After the skeleton on the first level of titles is defined,
one should develop the skeleton on the paragraph level;
this means defining all subtitles on lower levels and the
contents of all paragraphs under each lowest-level sub-
title. Finally, the last thing to do is to specify the first
sentence of each paragraph, which is the major one;
other sentences of each paragraph are just to explain
and/or justify the statement conveyed by the first
sentence.
It is not before now that the writing can start, and it
will be easy to do it; also, this approach enables that,
after the complete skeleton is developed by a senior
person (e.g., a major professor), the writing can be done
by a junior person (e.g., a graduate student); any errors
in writing will be localized at the paragraph level, and,
as such, easy to fix.
The above applies to research papers. An important
prerequisite for a good research paper is that a good
survey paper is prepared first, to demonstrate that major
solutions for the problem of interest are known.
In the case of a survey paper, the major requirement is
to have a concepts part (to define the major issues), and
the systems part (to define various algorithms and/or
implementations, etc.). The concepts part should be
preceded by a classification of concepts. The systems
part should be preceded by a classification of systems.
Each system in the systems part should be
IEEE TCCA NEWSLETTER, September 1996.
described/explained using the same template (e.g.,
origin, environment, essence, advantages, drawbacks,
relevant details, performance consideration, complexity
consideration, conclusion, trends, etc.). The choice of
elements for the template is flexible. What is not flexible
is that the same elements must be used in each template.
6. Semantics-Based
Layout of Transparencies
Major rules for doing the transparencies can be found
in numerous books. Consequently, the stress here is on
an issue which is extremely important, yet not mentioned
in any of the books known to this author - the rule about
the semantics-based layout of transparencies. This rule
reads as follows.
If a semantic entity must be spread over several lines,
the breakdown of lines should be done in a semantic
way. In other words, if a "bullet" is to be spread over
more than one line (often times, three is the maximum
which makes a good choice), each line should represent
a separate thought.
As an illustration, two examples are shown next, one
without and one with semantic splitting.
**********************************************
AN EXAMPLE WITHOUT SEMANTIC SPLITTING:
TOPIC TITLE
• Fixed/variable allocation scenarios based on the
home property (page manager): DSM + DSIO
• Writes get satisfied on distance or locally, depending
on what brings better performance
• Good if reads and writes are interleaved with similar
probabilities of occurrence
AN EXAMPLE WITH SEMANTIC SPLITTING:
TOPIC TITLE
• Fixed/variable allocation scenarios,
based on home property (page manager):
DSM + DSIO
• Writes get satisfied on distance or locally,
depending on what brings better performance
• Good if reads and writes are interleaved,
with similar probabilities of occurrence
**********************************************
4
 In other words, do not let the word-processor split the At last, but not least, the author is thankful to the
lines for you. Instead, do it by yourself, the right way! graduate students of the University of Belgrade for their
Semantic splitting is extremely useful for the audience, continuous efforts to educate their major professor, so he
and its fast comprehension of the material. An can keep up with the newest trends in the field (e.g.,
experiment was performed by the author to prove that Jovanka Ciric, Goran Davidovic, Ilija Ekmecic,
fact. In this experiment, the same subject was taught to Aleksandar Janicijevic, Milan Jovanovic, Aleksandar
two different groups of students, using two sets of Milenkovic, Zvezdan Petkovic, Milena Petrovic, Jelica
transparencies, one with and one without semantic Protic, and Dejan Raskovic).
splitting of lines. A test would be given after the subject
is completed. The experiment was repeated enough 10. References
times, and the test results were considerably different, in [Ekmecic97] Ekmecic,I., Tartalja,I., Milutinovic,V.,
favor of the transparencies based on semantic splitting. "Tutorial on Heterogeneous Processing:
Concepts and Systems,"
As a consequence of this experiment, the author of
IEEE Computer Society PRESS,
this paper insisted that transparencies for his university Los Alamitos, California, 1997.
courses and pre-conference tutorials are based on
[Flynn95] Flynn,M.J., "Computer Architecture,"
semantic splitting [Ekmecic97, Protic96, Tartalja96, Jones and Bartlett, Boston, Massachusetts,
Tomasevic93]. 1995.
Sometimes, semantic splitting seems impossible to do; [Hennessy96] Hennessy,J.L., Patterson,D.A.,
however, in each such case, it turns out that an "Computer Architecture:
alternative way of expressing the thoughts is both easy to A Quantitative Approach,"
Morgan Kaufmann,
split and sounds much better. San Francisco, California, 1996.
[Milutinovic95] Milutinovic,V.,
7. Conclusion "A Research Methodology in the Field of
This paper sets a standard for organization of research Computer Engineering for VLSI,"
Proceedings of the IEEE International
presentations, and defines the semantics-based layout of Conference on Microelectronics,
presentation transparencies. So far, almost without Nis, Serbia, Yugoslavia, September 1995.
exception, others would start using the views expressed [Milutinovic96] Milutinovic,V., "Surviving the Design of a
here (especially the semantics-based splitting for 200 MHz RISC Microprocessor:
transparencies), as soon as they learn about them, which Lessons Learned," IEEE CS PRESS,
Los Alamitos, California, 1996.
was a great source of pleasure and satisfaction for the
author. [Milutinovic97] Milutinovic,V., "Surviving the Design of a
16 MPs Multiprocessor System:
Lessons Learned," IEEE CS PRESS,
8. A Note Los Alamitos, California, 1997.
An earlier but wider version of this text can be found [Patterson94] Patterson,D.A., Hennessy,J.L., "Computer
in [Milutinovic95]. For lower level details, the interested Organization and Design," Morgan
Kaufmann, San Francisco, California, 1994.
reader is welcome to contact the author directly.
[Protic96] Protic,J., Tomasevic,M., Milutinovic,V.,
9. Acknowledgments "Tutorial on Distributed Shared Memory:
Concepts and Systems," IEEE CS PRESS,
The author is thankful to professors Mike Flynn of Los Alamitos, California, 1996.
Stanford and Yale Patt of Michigan for their response to [Tartalja96] Tartalja,I., Milutinovic, V., "Tutorial on
some of the ideas presented here; also, to professors Cache Coherency Maintenance in Shared
Memory Multiprocessors:
Hennessy, Gupta, and their graduate students for Software Solutions," IEEE CS PRESS,
numerous comments during the author’s last seminar at Los Alamitos, California, 1996.
Stanford University. Also, to professor Jean-Loup Baer [Tomasevic93] Tomasevic,I., Milutinovic, V., "Tutorial on
for his suggestion to publish a paper of this type with all Cache Coherency Maintenance in Shared
the experiences incorporated, and to professor Nitin Memory Multiprocessors:
Hardware Solutions," IEEE CS PRESS,
Vaidya for his efforts to help about the paper quality.
Los Alamitos, California, 1993.

IEEE TCCA NEWSLETTER, September 1996. 5

Epilog
This section includes a list with some of the author’s
journal papers which either helped create the research
and presentation methodology which is the subject of
this paper, or are based on the research and presentation
methodology presented in this paper. The enclosed list
includes only the papers published after 1.1.1990. and
only from the prestigeous IEEE periodicals.
1. V. Milutinovic,
“Mapping of Neural Networks
onto the Honeycomb Architecture,”
Proceedings of the IEEE , Vol. 77, No. 12, December 1990.
2. V. Milutinovic,
“Tutorial on Microprogramming
and Firmware Engineering,”
IEEE Computer Society Press, Los Alamitos, California,
1990.
3. B. Perunicic, S. Lakhani, V. Milutinovic,
“Stochastic Modeling and Analysis
of Propagation Delays in GaAs Adders,”
IEEE Transactions on Computers, Vol. 40, No. 1,
January 1991.
4. V. Milutinovic, D. Fura, W. Helbig,
“Pipeline Design Trade-offs
in 32-bit Gallium Arsenide Microprocessor,”
IEEE Transactions on Computers, Vol. 40, No. 11,
November 1991.
5. L. Hoevel, V. Milutinovic,
“Terminology Risks with the RISC Concept
in the Risky RISC Arena,”
IEEE Computer, Vol. 25, No. 1, January 1992
(Open Channel).
6. M. Tomasevic, V. Milutinovic,
“Tutorial on the Cache Coherency Problem
in Shared-Memory Multiprocessors: Hardware Solutions,”
IEEE Computer Society Press, Los Alamitos, California,
1993.
7. M. Tomasevic, V. Milutinovic,
“A Survey of Hardware Solutions for Maintenance
of Cache Consistency
in Shared Memory Multiprocessor Systems,”
IEEE MICRO (Part #1), October 1994.
8. M. Tomasevic, V. Milutinovic,
“A Survey of Hardware Solutions for Maintenance
of Cache Consistency
in Shared Memory Multiprocessor Systems,”
IEEE MICRO (Part #2), December 1994.
9. V. Milutinovic, Z. Petkovic,
“Processor Design Using Silicon Compilation:
Ten Lessons Learned from a RISC Design,”
IEEE Computer, Vol. 28, No. 3, March 1995
(Open Channel).
10. S. Savic, M. Tomasevic, V. Milutinovic,
“Improved RMS for the PC Environment,”
Microprocessor Systems, Vol. 19, No. 10, September 1995
(A follow up paper will be published in an IEEE journal).
11. I. Ekmecic, I. Tartalja, V. Milutinovic,
“A Taxonomy of Heterogeneous Computing,”
IEEE Computer, Vol. 28, No. 12, December 1995
(Hot Topics).
IEEE TCCA NEWSLETTER, September 1996.
12. I. Tartalja, V. Milutinovic,
“Tutorial on the Cache Coherency Problem
in Shared-Memory Multiprocesors: Software Solutions,”
IEEE Computer Society Press, Los Alamitos, California,
1996.
13. M. Tomasevic, V. Milutinovic,
“The World Invalidate Protocol,”
Microprocessor Systems, January 1996
(A follow up paper will be published in an IEEE journal).
14. A. Grujic, M. Tomasevic, V. Milutinovic,
“A Simulation Study of Hardware DSM Approaches,”
IEEE Parallel and Distributed Technology, Spring 1996.
15. D. Milutinovic, V. Milutinovic,
“Mapping of Interconnection Networks
for Parallel Processing onto the Sea-of-Gates VLSI,”
IEEE Computer, Vol. 29, No. 4, April 1996.
16. J. Protic, M. Tomasevic, V. Milutinovic,
“A Survey of Distributed Shared Memory:
Concepts and Systems,”
IEEE Parallel and Distributed Technology, Summer 1996.
17. I. Tartalja, V. Milutinovic,
“A Survey of Software Solutions for Cache Consistency
Maintenance in Shared Memory Multiprocessors,”
IEEE Software, Fall 1996.
18. V. Milutinovic,
“Surviving the Design of a 200MHz RISC Microprocessor:
Lessons Learned,”
IEEE Computer Society Press, Los Alamitos, California,
1996.
19. J. Protic, M. Tomasevic, V. Milutinovic,
“Tutorial on DSM: Concepts and Systems,”
IEEE Computer Society Press, Los Alamitos, California,
1996.
20. I. Ekmecic, I. Tartalja, V. Milutinovic,
“A Survey of Heterogeneous Computing:
Concepts and Systems,”
Proceedings of the IEEE, August 1996.
21. D. Milicev, Z. Petkovic, D. Raskovic, D. Jelic, D.
Jelisavcic, D. Stevanovic, V. Milutinovic,
“Modeling of Modern 32-bit and 64-bit Microprocessors,”
IEEE Transactions on Education, 1996.
22. V. Milutinovic,
“The Best Method for Presentation of Research Results
in Computer Engineering,”
IEEE TCCA Newsletter, September 1996.
23. V. Milutinovic,
“Some Solutions for Critical Problems
of Distributed Shared Memory Systems:
New Ideas to Analyse,”
IEEE TCCA Newsletter, September 1996.
24. V. Milutinovic, M. Tomasevic, B. Markovic, M. Tremblay,
“The Split Temporal/Spatial Cache Memory
for Next Generation SuperMicroprocessors,”
(To be published).
Conference version available from the Proceedings of the
IEEE SCIzzL-5, Santa Clara, California, March 1996.
25. V. Milutinovic, A. Milenkovic, J. Ristic, G. Shaeffer,
“The Direct Injection/Replacement Cache Memory
for Next Generation SuperMicroprocessors,”
(To be published).
Conference version available from the Proceedings of the
IFACT Encore-96, Belgrade, Yugoslavia, July 1996.
6